August 2013 – ICCL Summer School Dresden – Pascal Hitzler
Ontology Design Patterns Modeling Examples
Pascal Hitzler Kno.e.sis Center
Wright State University, Dayton, OH http://www.knoesis.org/pascal/
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 2
EarthCube
NSF effort for the earth sciences Goal: To transform the conduct of research in the geosciences by developing IT solutions for the integration of information and data in the geosciences. How this is going to be done is still in the making. Semantic Technologies have been part of the mix from the start. [Berg-Cross, …, Hitzler et al., GIBDa 2012]
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 3
EarthCube requires
• information integration • interoperability • conceptual
modeling • intelligent
search • data-model
intercomparison • data publishing
support
Semantic Web studies
• information integration • interoperability • conceptual
modeling • intelligent
search • data-model
intercomparison • data publishing
support Pascal Hitzler, WSU; Krzysztof Janowicz, UCSB
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 4
Vertical data integration
Query Upper level ontology
Dataset
Dataset
Answer
[Joshi, Jain, Hitzler et al. ODBASE 2012]
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 5
Ontological commitments
Two ontologies. Left: transportation domain Right: agriculture domain We cannot simply equate a:Canal and b:Canal !
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 6
Linked Data: Variety and Value (GovTrack)
“Nancy Pelosi voted in favor of the Health Care Bill.”
Bills:h3962
H.R. 3962: Affordable Health Care for America
Act
Votes:2009-887/+
people/P000197
Nancy Pelosi On Passage: H R 3962 Affordable Health Care for
America Act
Vote: 2009-887
vote:hasAction
vote:vote
dc:title
vote:hasOption
rdfs:label Aye
dc:title
vote:votedBy
name
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 7
Ontology Design Patterns
• Bottom-up homogenization of data representation.
• Avoidance of strong ontological commitments.
• Avoidance of standardization.
• Well thought-out patterns can be very strong and versatile, thus serve many needs.
We are currently establishing many geo-patterns in a series of
hands-on workshops, the GeoVoCamps, see http://vocamp.org/
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 8
Ontology Design Patterns
Pattern1 Pattern1
Pattern2 Pattern2
Pattern2
Pattern3
Pattern3
“Horizontal” alignment via patterns
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 9
Patterns TOC
• Semantic Trajectories • Biodiversity • Map Scaling • Part-of Relationships
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 10
Semantic Trajectories
[Hu, Janowicz, Carral, Scheider, Kuhn, Berg-Cross, Hitzler, Dean, COSIT2013, to appear]
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 11
Semantic Trajectories
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 12
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 13
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 14
Patterns TOC
• Semantic Trajectories • Biodiversity • Map Scaling • Part-of Relationships
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 15
Type counting
[ACM GIS 2012]
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 16
Non-monotonicity
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 17
Semantics
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 18
Closure
• Straightforward carrying over of circumscription to DLs: undecidable for expressive DLs [Bonatti, Lutz, Wolter, KR2006, JAIR 2009] Unintuitive modeling: extensions of minimized predicates may contain unknown individuals
• Fixing the unintuitive aspect: allow only named individuals in extensions of minimized predicates decidable even for very expressive DLs we also have a tableaux algorithm [Sengupta, Krisnadhi, Hitzler, ISWC2011] called Grounded Circumscription
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 19
Circumscription
• Use a knowledge base K as usual. • Additionally, specify “circumscribed” (minimized) predicates.
• Among all models M of K, the circumscribed models (c-models)
are those for which there is no model which is preferred over M. A model J is preferred over M if a) they have the same domain of discourse b) constants have the same extensions in both models c) the J-extension of each minimized predicate is contained in its M-extension d) the J-extension of some minimized predicate is strictly contained in its M-extension
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 20
Grounded Circumscription for DLs
• Use a knowledge base K as usual. • Additionally, specify “circumscribed” (minimized) predicates.
• Among all models M of K, the circumscribed models (gc-models)
are those for which there is no model which is preferred over M and extensions of minimized predicates contain only named individuals. A model J is preferred over M if a) they have the same domain of discourse b) constants have the same extensions in both models c) the J-extension of each minimized predicate is contained in its M-extension d) the J-extension of some minimized predicate is strictly contained in its M-extension
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 21
Circumscription vs. Grounded Circ.
• Circumscription: – minimization of roles leads to undecidability (for non-empty
Tboxes
• Grounded Circumscription: – Decidable even under role grounding for very expressive
decidable DLs. – Complexity upper bound for satisfiability or for finding a gc-
model is EXPC, where C is the complexity of the underlying DL.
We also have a tableaux algorithm for different reasoning tasks.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 22
Example
Both of are not logical consequences under classical DL semantics. However, they are logical consequences when hasAuthor is
minimized (using the UNA).
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 23
Patterns TOC
• Semantic Trajectories • Biodiversity • Map Scaling • Part-of Relationships
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 24
Cartographic Map Scaling
[Carral, Scheider, Janowicz, Vardeman, Krisnadhi, Hitzler, ESWC2013]
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 25
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 26
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 27
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 28
Semantics in OWL
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 29
Patterns TOC
• Semantic Trajectories • Biodiversity • Map Scaling • Part-of Relationships
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 30
Source
Content taken from Morton E. Winston, Roger Chaffin, Douglas Herrmann, A Taxonomy
of Part-Whole Relations, Cognitive Science 11, 417-444, 1987. and the OWL modeling from Prateek Jain, Pascal Hitzler, Kunal Verma, Peter Yeh, Amit Sheth,
Moving beyond sameAs with PLATO: Partonomy detection for Linked Data. In: Ethan V. Munson, Markus Strohmaier (Eds.): 23rd ACM Conference on Hypertext and Social Media, HT '12, Milwaukee, WI, USA, June 25-28, 2012. ACM, 2012, pp. 33-42.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 31
part-of relationships
• the X is part of the Y • X is partly Y • X’s are part of Y’s • X is a part of Y • The parts of a Y include the Xs, the Zs, …
• The head is part of the body • Bicycles are partly aluminum • Pistons are part of engines • Dating is a part of adolescence • The parts of a flower include the stamen, the petals, etc. …
• “meronymic” relations (“meros” is greek for “part”)
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 32
part-of: a possible view
One could think that part-of is a binary relation which is • a strict partial ordering, i.e.
– transitive If X part of Y, and Y part of Z. Then X part of Z.
– irreflexive X is never part of X.
– antisymmetric If X part of Y. Then Y is never part of X.
However, this view is problematic.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 33
Transitivity
Simpson’s finger is part of Simpson’s hand. Simpson’s hand is part of Simpson’s body. Simpson’s finger is part of Simpson’s body. This works, but the following doesn’t: Simpson’s arm is part of Simpson. Simpson is part of the Philosophy Department. Hence(?) Simpson’s arm is part of the Philosophy Department. So when do we have transitivity?
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 34
Winston’s approach
Distinguish 6 different types of meronymic relations: 1. component – integral object (pedal – bike) 2. member – collection (ship – fleet) 3. portion – mass (slice – pie) 4. stuff – object (steel – car) 5. feature – activity (paying – shopping) 6. place – area (Everglades – Florida)
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 35
Dimensions of meronymic relations
A type of part-of relationships • functional
Functional parts are restricted, by their function, in their spatial or temporal location. handle – cup
• homeomerous Homeomerous parts are the same kind of thing as their wholes. slice – pie but not tree – forest
• separable Separable parts can in principle be separated from the whole. handle – cup but not steel – bike
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 36
Dimensions
From Winston et al., A Taxonomy of Part-whole Relations, Cognitive Science 11, 417-444, 1987.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 37
Component – Integral Object
• A handle is part of a cup. • Wheels are parts of cars. • The refrigerator is part of the kitchen. • Chapters are parts of books. • A punchline is part of a joke. • Belgium is part of NATO. • Phonology is part of linguistics. • The viola part in a symphony.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 38
Member – Collection
• A tree is part of a forest. • A juror is part of a jury. • This ship is part of a fleet. Do not confuse with class – member relationships, such as • The Nile is a river. • Fido is a dog. which are not part-of relationships. class membership: determined on the basis of similarity to other
members. member – collection: determined on the basis of spatial proximity
or by social connection.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 39
Portion – Mass
• This slice is part of a pie. • A yard is part of a mile. • This hunk is part of my clay.
Homeomerous: Every portion of a pie is “pie”. (while, e.g., a window is quite unlike the house of which it is part.)
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 40
Portion – Mass
Can be distinguished from component – integral object by substituting the phrase “some of”:
• She asked me for part of my orange. (… for some of my orange)
However *not*: The engine is some of the car. This test won’t distinguish from member – collection: • Some of the fraternity brothers are sophomores.
(this is the “count” sense of “some”, not the “mass” sense) However, for member – collection we can phrase it as: • One of the brothers is a sophomore.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 41
Stuff – Object
• A martini is partly alcohol. • The bike is partly steel. • Water is partly hydrogen.
By asking for: “What is it made of?” (For component – integral object we would ask:
“What are its parts?”) Stuff cannot be separated from the object.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 42
Feature – activity
• Paying is part of shopping. • Bidding is part of playing Bridge. • Ovulation is part of the menstrual cycle. • Dating is part of adolescence.
Features or phases of activities and processes. Unlike the other types, in this case we cannot say “X has Y”, such
as for others in • Sororities have members. • Bicycles have pedals • Plays have acts. E.g. we cannot say “Shopping has paying”.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 43
Place – Area
• The Everglades are part of Florida. • An oasis is a part of a desert. • The baseline is part of a tennis court.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 44
Other apparently similar relations which are not meronymic
• Topological Inclusion – The wine is in the cooler. – The meeting is in the morning. – Careful: “The Everglades are part of Florida” is meronymic.
But “West Berlin is part of East Germany” is wrong. [Note paper was written 1987.]
• Class Inclusion – Cars are a type of vehicle. – Theft is a crime. – Careful: “Frying is a type of cooking” is meronymic, as is
“Honesty is a type of virtue”.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 45
Other apparently similar relations which are not meronymic
• Attribution – Towers are tall. – Coal burns. – The joke was funny.
• Attachment – Earrings are attached to ears. – Fingers are attached to hands.
(note: they are also parts of hands) • Ownership
– A millionaire has money. – The author has the copyright. – Jenny has a bicycle.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 46
Transitivity again
Simpson’s finger is part of Simpson’s hand. Simpson’s hand is part of Simpson’s body. Simpson’s finger is part of Simpson’s body. This works, but the following doesn’t: Simpson’s arm is part of Simpson. Simpson is part of the Philosophy Department. Hence(?) Simpson’s arm is part of the Philosophy Department. Winston argues: If we combine two sentences with the same type
of meronymic relation, then we have transitivity. Indeed, in all mixed cases, counterexamples to transitivity can be found (given in the paper).
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 47
Other properties – and some OWL modeling
Winston et al. list several properties of meronymic relations. First some notation for the 6 types of part-of relations: • po-component • po-member • po-portion • po-stuff • po-feature • po-place PO is the set containing these six binary relations. • part-of: The “general” part-whole relation. • spatially-located-in: topological located-in relationship
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 48
1. For all R 2 PO, R is transitive, asymmetric, and irreflexive (i.e., a strict partial order).
2. For all R 2 PO, R v part-of. Does not imply transitivity of part-of.
3. spatially-located-in is transitive and reflexive. 4. For all R 2 PO, we have
– R ± spatially-located-in v spatially-located-in – spatially-located-in ± Rv spatially-located-in
5. For all R 2 PO [ {spatially-located-in} and all classes C, we have (8x)(8y)(R(x,y)ÆC(y) → (9z)(R(x,z)ÆC(z))).
6. For all R 2 PO [ {spatially-located-in} and all classes C, we have (8x)(8y)(C(y) Æ (C(y) → R(x,y)) → R(x,y)).
Note: 5+6 are tautologies, so need not be modeled in OWL.
Axioms (extracted from Winston et al.)
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 49
Meronymic relations in OWL
1. For all R 2 PO, R is transitive, asymmetric, and irreflexive (i.e., a strict partial order).
2. For all R 2 PO, R v part-of. Does not imply transitivity of part-of.
3. spatially-located-in is transitive and reflexive. 4. For all R 2 PO, we have
– R ± spatially-located-in v spatially-located-in – spatially-located-in ± Rv spatially-located-in
This results in a total of 3¢6+2¢6+2+6¢2 = 44 axioms, all expressible
in OWL 2. However, there is a catch!
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 50
A Catch
1. For all R 2 PO, R is transitive, asymmetric, and irreflexive (i.e., a strict partial order).
Problem: A relation in OWL 2 DL cannot be transitive and reflexive at the same time: A transitive property is complex, and thus not simple. However only simple properties are allowed to be irreflexive.
So: this ends up in OWL 2 Full. Straightforward fix:
Drop irreflexivity. This will probably work in most cases. Better fixes are based on rules or nominal schemas (covered later
in class).
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 51
Another two catches
All properties occuring in the above given part-of ontology are complex (i.e., non-simple).
OWL 2 has global restrictions on the use of such properties. This hampers modeling, and may yield to OWL 2 Full ontologies
after all desired relationships have been modeled. Another problem: Regularity conditions may become violated if
merging the part-of ontology with a domain ontology. Fixes: as above (drop some axioms) Better: rules or nominal schemas (covered later in class).
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 52
Addressing the issues
We have several issues with modeling the part-of ontology following Winston.
E.g., relations cannot be transitive, asymmetric, and irreflexive at
the same time. We can now approximate this as follows: Characterize the relation (e.g., R) as transitive and asymmetric. Furthermore, specify {x} u 9R.{x} v?.
More generally, if you run into a rule which you cannot model in
OWL, simply approximate using nominal schemas.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 53
References
• Pascal Hitzler, Krzysztof Janowicz, Linked Data, Big Data, and the 4th Paradigm. Semantic Web 4 (3), 2013, 233-235.
• Krzysztof Janowicz, Pascal Hitzler, The Digital Earth as Knowledge Engine. Semantic Web 3 (3), 213-221, 2012.
• Gary Berg-Cross, Isabel Cruz, Mike Dean, Tim Finin, Mark Gahegan, Pascal Hitzler, Hook Hua, Krzysztof Janowicz, Naicong Li, Philip Murphy, Bryce Nordgren, Leo Obrst, Mark Schildhauer, Amit Sheth, Krishna Sinha, Anne Thessen, Nancy Wiegand, Ilya Zaslavsky, Semantics and Ontologies for EarthCube. In: K. Janowicz, C. Kessler, T. Kauppinen, D. Kolas, S. Scheider (eds.), Workshop on GIScience in the Big Data Age, In conjunction with the seventh International Conference on Geographic Information Science 2012 (GIScience 2012), Columbus, Ohio, USA. September 18th, 2012. Proceedings.
• Krzysztof Janowicz, Pascal Hitzler, Thoughts on the Complex Relation Between Linked Data, Semantic Annotations, and Ontologies. In: Proceedings of the CIKM 2013 Workshop on Exploiting Semantic Annotation, ESAIR 2013. To appear.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 54
References
• Pascal Hitzler, Frank van Harmelen, A reasonable Semantic Web. Semantic Web 1 (1-2), 39-44, 2010.
• Prateek Jain, Pascal Hitzler, Peter Z. Yeh, Kunal Verma, Amit P. Sheth, Linked Data is Merely More Data. In: Dan Brickley, Vinay K. Chaudhri, Harry Halpin, Deborah McGuinness: Linked Data Meets Artificial Intelligence. Technical Report SS-10-07, AAAI Press, Menlo Park, California, 2010, pp. 82-86. ISBN 978-1-57735-461-1. Proceedings of LinkedAI at the AAAI Spring Symposium, March 2010.
• Amit Krishna Joshi, Prateek Jain, Pascal Hitzler, Peter Z. Yeh, Kunal Verma, Amit P. Sheth, Mariana Damova, Alignment-based Querying of Linked Open Data. In: Meersman, R.; Panetto, H.; Dillon, T.; Rinderle-Ma, S.; Dadam, P.; Zhou, X.; Pearson, S.; Ferscha, A.; Bergamaschi, S.; Cruz, I.F. (eds.), On the Move to Meaningful Internet Systems: OTM 2012, Confederated International Conferences: CoopIS, DOA-SVI, and ODBASE 2012, Rome, Italy, September 10-14, 2012, Proceedings, Part II. Lecture Notes in Computer Science Vol. 7566, Springer, Heidelberg, 2012, pp. 807-824.
• Yingjie Hu, Krzysztof Janowicz, David Carral, Simon Scheider, Werner Kuhn, Gary Berg-Cross, Pascal Hitzler, Mike Dean, A Geo-Ontology Design Pattern for Semantic Trajectories. In: Proceedings of COSIT 2013, Conference on Spatial Information Theory, Scarborough, UK, September 2013. To appear.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 55
References
• David Carral Martinez, Krzysztof Janowicz, Pascal Hitzler, A Logical Geo-Ontology Design Pattern for Quantifying over Types. In: Isabel F. Cruz, Craig Knoblock, Peer Kröger, Egemen Tanin, Peter Widmayer (Eds.): SIGSPATIAL 2012 International Conference on Advances in Geographic Information Systems (formerly known as GIS), SIGSPATIAL'12, Redondo Beach, CA, USA, November 7-9, 2012. ACM 2012, pp. 239-248.
• David Carral, Simon Scheider, Krzysztof Janowicz, Charles Vardeman, Adila A. Krisnadhi, Pascal Hitzler, An Ontology Design Pattern for Cartographic Map Scaling. In: Philipp Cimiano, Oscar Corcho, Valentina Presutti, Laura Hollink, Sebastian Rudolph (Eds.), The Semantic Web: Semantics and Big Data. 10th International Conference, ESWC 2013, Montpellier, France, May 26-30, 2013. Proceedings. Lecture Notes in Computer Science Vol. 7882, Springer, Heidelberg, 2013, pp. 76-93.
• Prateek Jain, Pascal Hitzler, Kunal Verma, Peter Yeh, Amit Sheth, Moving beyond sameAs with PLATO: Partonomy detection for Linked Data. In: Ethan V. Munson, Markus Strohmaier (Eds.): 23rd ACM Conference on Hypertext and Social Media, HT '12, Milwaukee, WI, USA, June 25-28, 2012. ACM, 2012, pp. 33-42.
August 2013 – ICCL Summer School Dresden – Pascal Hitzler 56
References
• Piero A. Bonatti, Carsten Lutz, Frank Wolter: The Complexity of Circumscription in DLs. J. Artif. Intell. Res. (JAIR) 35: 717-773 (2009)
• Kunal Sengupta, Adila Krisnadhi, Pascal Hitzler, Local Closed World Reasoning: Grounded Circumscription for OWL. In: L. Aroyo, C. Welty, H. Alani, J. Taylor, A. Bernstein, L. Kagal, N. F. Noy, E. Blomqvist (Eds.): The Semantic Web – ISWC 2011 - 10th International Semantic Web Conference, Bonn, Germany, October 23-27, 2011, Proceedings, Part I. Lecture Notes in Computer Science Vol. 7031, Springer, Heidelberg, 2011, pp. 617-632.
• Markus Krötzsch, Frederick Maier, Adila Alfa Krisnadhi, Pascal Hitzler, A Better Uncle For OWL – Nominal Schemas for Integrating Rules and Ontologies. In: S. Sadagopan, Krithi Ramamritham, Arun Kumar, M.P. Ravindra, Elisa Bertino, Ravi Kumar (eds.), WWW '11 20th International World Wide Web Conference, Hyderabad, India, March/April 2011. ACM, New York, 2011, pp. 645-654.
Top Related